Message boards :
Number crunching :
different credits for same unit - an unusual aspect of it
Message board moderation
Author | Message |
---|---|
Send message Joined: 24 Oct 04 Posts: 18 Credit: 1,485 RAC: 0 |
http://lhcathome.cern.ch/workunit.php?wuid=359639 I wonder why 3 people got credited 49.84 for this unit and other 3 got 99.69 I thought only 3 results were needed and that everybody would get the same number of credits... Quite puzzled... |
Send message Joined: 17 Sep 04 Posts: 10 Credit: 5,620 RAC: 0 |
If I remember correct they announced some time ago that the credit-system was goning to change here at LHC. You should get credit based on how well your result match the others. Close enough and you get full credit, a bit of and you get less and so on... Something like that anyway. It's better for us to get some credit even if the result don't match completley because we did crunch and dedicate CPU-time for the project. Look under the FAQ under 2.5 How and when do I get credit? |
Send message Joined: 24 Oct 04 Posts: 18 Credit: 1,485 RAC: 0 |
Thanks a lot, that was really helpful! |
Send message Joined: 17 Sep 04 Posts: 49 Credit: 25,253 RAC: 0 |
Got to do more reading........I hate giving bad info out...... |
Send message Joined: 1 Sep 04 Posts: 275 Credit: 2,652,452 RAC: 0 |
> If I remember correct they announced some time ago that the credit-system was > goning to change here at LHC. > You should get credit based on how well your result match the others. > Close enough and you get full credit, a bit of and you get less and so on... > > Something like that anyway. > It's better for us to get some credit even if the result don't match > completley because we did crunch and dedicate CPU-time for the project. > > Look under the FAQ under > 2.5 How and when do I get credit? > I also remember this announcement. It also seems like they mentioned that late work could get partial credit, as long as it was returned before the database entry was cleared. That may be why some are getting zero when it seems like they should be getting something different. John Keck -- BOINCing since 2002/12/08 -- |
Send message Joined: 3 Oct 04 Posts: 7 Credit: 55,484 RAC: 0 |
The linux BOINC client is horribly misoptimized*. Therefore, benchmarks are lower than they should be. This brings the entire credit value down; TIME TO CRUNCH WU X BENCHMARK VALUES = CLAIMED CREDIT The lower claimed credit brings the granted credit down too; the granted credit is the median of claimed credits(used to be the lowest). *Although the Boinc client is misoptimized, most boinc exceutables(number crunchers) are not. |
Send message Joined: 27 Sep 04 Posts: 34 Credit: 199,100 RAC: 0 |
> The linux BOINC client is horribly misoptimized*. Actually it's the windows client benchmark code that is overoptomized. This is being fixed in the coming BOINC 4.5x. |
Send message Joined: 2 Sep 04 Posts: 121 Credit: 592,214 RAC: 0 |
> Actually it's the windows client benchmark code that is overoptomized. This is > being fixed in the coming BOINC 4.5x. Ever since I'm using the gcc 3.4.3 optimized BOINC 4.13 Core, my Linux Boxes score about equal to my few Windows machines. Therefor, I'd say the Benchmark of the Linux BOINC 4.13 is indeed terribly un-optimized. (The RAM-Benchmark is completely un-optimized across all Platforms for example, maybe they fix that as well sometime in the future) Optimized Linux BOINC 4.xx Downloads (precompiled, ready-to-use) : http://boinc.us.tt/ or http://www.pperry.f2s.com/ ___________________________________________ <p>Scientific Network : 38322 MHz �� 9484 MB �� 895.4 GB </p> |
©2024 CERN